The US Has Begun Discussing Sanctions Against China And Russia For Copying US AI

Republicans intend to crack down on theft.

Republicans in the US House of Representatives have proposed an initiative to impose sanctions against companies that illegally copy advanced American artificial intelligence models to create their own competing systems. The relevant bill, dubbed the "Preventing the Theft of American AI Models Act," was introduced by Congressman Bill Huizenga with the support of John Moolenar, chairman of the House Select Committee on China. The document will be considered next week in the Foreign Affairs Committee, reports Bloomberg.

The bill requires the U.S. government to identify entities in China and Russia that use methods to unauthorizedly copy and extract data from U.S. AI models. Violators are proposed to be sanctioned through a Commerce Department blacklist, as well as through the president's emergency economic powers under a 1977 law. Potential targets include Chinese labs DeepSeek, Moonshot, and MiniMax, among others, as well as larger players such as Alibaba and ByteDance.

"Attacks to extract data from models are the last frontier of Chinese economic coercion and theft of American intellectual property," Huizenga said. He emphasized that U.S. AI models demonstrate breakthrough capabilities, and it is crucial to prevent their theft.

The initiative was prompted by concerns of leading AI developers, including OpenAI, Anthropic and Google (Alphabet Inc.), which have recorded cases of creating cheaper and less secure copies of their products. This is so-called "adversarial distillation," a technique in which a newer model is trained on an existing model, replicating its functionality at a much lower cost. While the use of such technology is allowed with restrictions, using it to copy cutting-edge designs without authorization violates company rules.

The report by U.S. lawmakers calls on the Commerce Department to begin treating such actions as industrial espionage. It also recommends that such cases be referred to the Justice Department for possible criminal prosecution.

OpenAI previously reported that DeepSeek improperly used ChatGPT results to create a fake model devoid of security mechanisms. Google and Anthropic also published reports with similar findings. U.S. AI labs have warned that "distilled" models are vulnerable to censorship on topics sensitive to the Chinese government and may not prevent potentially dangerous user actions - for example, when searching for information about the creation of biological weapons.

In addition, open-source models created by Chinese startups undermine the revenue needed for U.S. companies to recoup investments in data centers and pay staff.

You can support the website Charter97.org

MULTI-CURRENCY ACCOUNT FOR ASSISTANCE:

Bank's name: Bank Millennium S.A.
Address: ul. Stanislawa Zaryna, 2A, 02-593, Warszawa
IBAN: PL97116022020000000216711123
SWIFT: BIGBPLPW
Name of the account holder: Fundacja “KARTA ‘97”
Purpose/title of payment: Donation for statuary aims

You can contact us by the e-mail [email protected]